Apple Core ML

Detects objects, faces and license plates in the video of your cameras on a Mac, using Apple CoreML, and provides the semantic search (CLIP).

What it does

• Detects people, vehicles and animals
• Detects faces and recognizes known people
• Detects license plates and reads their text
• Provides the CLIP semantic search for events
• Follows the confidence values from the camera's detection settings
• Lets you choose which parts of the Apple hardware do the work: CPU, GPU and Neural Engine
• Downloads its models itself

What you need

• A ViON server running on a Mac (macOS)
• For names on faces, this plugin must be selected as "Face Recognition" for each camera: camera settings, "Plugins", "Detections"

Settings

• CLIP model (images): the model for the semantic search, shared by all cameras. After a change the recordings have to be reindexed
• Face recognition model: shared by all cameras. After a change the saved faces are recalculated
• Compute units: where CoreML runs the models. ALL uses CPU, GPU and Neural Engine; the other options use only some of them
• Active hardware: shows where the models are running
• Download models again: deletes the downloaded models and downloads the current ones
• Reset settings: puts all plugin settings back to their defaults
• Per camera, in the groups "Object detection", "Face detection" and "License plates": Model, Detection model and OCR model. "Default" follows the recommended model